Women without degrees saw weaker employment growth in AI-exposed work. AI is not yet the proven cause.

What the comparison found

Across 480 occupations, each one-point increase in Pew’s AI-exposure score corresponded to 6.3% employment growth for men and 0.2% for women. For workers with a high school diploma or less, the corresponding figures were +8.8% for men and −9.5% for women. The score runs from zero to 3.5 and describes tasks AI could assist with or perform, not AI tools used by employers.

For workers with bachelor’s degrees or more, neither the male nor female exposure gradient was statistically distinct from zero. The result does not mean every highly exposed occupation lost jobs: Pew reports that total employment in the most exposed occupations grew.

Why the result does not prove displacement

Pew’s technical appendix uses CPS employment counts by occupation, year and sex, with occupation and year controls and pre-pandemic employment weights. Its exposure score is built from O*NET work activities. The pre/post comparison cannot separate AI from other changes in hiring, industry demand or occupational composition. Men without a degree are also relatively uncommon in the most exposed occupations, so the male subgroup result should not be read as a universal advantage.

What would narrow the answer

Records linking actual employer AI adoption to workers’ hires, exits and employment by gender and education could test whether this gradient tracks AI use. Until then, the new subgroup warning can coexist with the younger-worker payroll decline and broad tracker’s lack of a clear economy-wide AI footprint: they study different groups and comparisons.